Delaware is considering legislation that would allow an Artificial Intelligence Company, or AIC, to establish and operate a business with decision-making handled by AI agents and only a single human involved. The proposal is attracting scrutiny from members of the US Congress, who are questioning how responsibility would be assigned if an AI-run company caused harm.
The state is already known for business-friendly rules, including favorable tax arrangements in some circumstances and specialized courts. Supporters of the proposal argue that it could help Delaware provide legal certainty for companies built around agentic AI and prevent such businesses from moving overseas.
Under the draft bill, an AI-run company could operate any lawful business except banking. It would initially enter a 30-month regulatory sandbox, during which the AI agent could make business decisions while the concept is tested. The company would be able to bring legal action and could also be sued like other businesses.
The proposal has drawn the attention of the House Oversight Committee. Its chairman, Republican Representative James Comer, wrote to Delaware Governor Matt Meyer requesting safety assessments. Comer warned that the framework could potentially protect powerful AI companies and their investors from responsibility for reckless or harmful actions. He also indicated that federal legislation could follow if the assessments do not produce satisfactory results.
Legal experts have raised concerns that the legislation could create an overly broad liability shield. Andrew Baker, a law professor at the University of California, told the Financial Times that the bill may be intended to establish legal certainty for Delaware-registered companies using AI agents, but said a generous liability shield would be difficult to justify without a clearly developed policy rationale.
The debate centers on the difference between conventional corporations and companies directed by AI. Corporate structures can limit personal liability, but traditional businesses are ultimately managed by people who remain accountable for their decisions. An AI agent has no equivalent personal or moral responsibility, leaving the bill to determine how liability would be handled when an autonomous system makes a harmful choice.
Sources: TechRadar.